RowSpeak Private Deployment FAQ

This document answers the questions we hear most often from IT teams, security reviewers, and procurement stakeholders evaluating RowSpeak Private Deployment.

Deployment and Infrastructure

Can RowSpeak be deployed on-premises?
Yes. RowSpeak runs on your own servers inside your data center or internal network. You control the hardware, the operating system, and the network configuration.

Can RowSpeak run in a private cloud?
Yes. RowSpeak can be deployed in your own AWS VPC, Azure VNet, or GCP VPC. You deploy it in your own account — RowSpeak does not have access to your cloud environment.

Does RowSpeak support air-gapped deployment (no internet)?
Yes. With open-source models, RowSpeak runs with no outbound internet connection. Model weights are pre-loaded during setup. All processing happens locally.

What operating systems are supported?
Ubuntu 22.04 LTS is the primary supported OS. RHEL 8+ is also supported. Windows Server is supported via Docker Desktop or WSL2.

Does RowSpeak support Docker?
Yes. Docker and Docker Compose are the standard deployment method. Kubernetes is also supported for high-availability and multi-node configurations.

How long does deployment take?
A standard single-server deployment typically takes 1–3 days, including environment setup, model loading, and initial testing. The RowSpeak team provides remote support throughout.

What are the minimum hardware requirements?
For a small team (up to 20 users): 8-core CPU, 32 GB RAM, NVIDIA GPU with 16 GB VRAM, 500 GB SSD. See the full hardware sizing guide for larger configurations.

Do we need a GPU? A GPU is required if you run open-source models locally. If you use closed-source models via API (GPT, Claude, Gemini), GPU requirements are significantly reduced.

Data Handling and Security

Does RowSpeak send our data to any external server?
No. With open-source models, all data stays inside your environment. RowSpeak does not transmit your files, outputs, or logs to any external server.

What if we use a closed-source model like GPT or Claude?
If you choose a closed-source model, your data is sent to that provider's API using your own API key and your own account. RowSpeak does not act as an intermediary. You are subject to that provider's data handling policies.

Does RowSpeak collect usage telemetry from our deployment?
No. RowSpeak Private Deployment does not send usage data, analytics, or telemetry back to RowSpeak.

Where are files stored after upload?
Files are stored in your own environment — on your server's local storage or a storage system you configure. RowSpeak does not retain copies externally.

How is data encrypted?
Data in transit between your users and the RowSpeak server uses TLS (you configure your own certificates). Data at rest is managed by your infrastructure — RowSpeak stores files in your designated storage, and encryption at rest is handled by your disk or volume encryption setup.

Does RowSpeak support role-based access control?
Yes. RowSpeak supports Admin, Editor, and Viewer roles. Workspace isolation is also supported, allowing different teams to have separate environments with separate access controls.

Does RowSpeak support SSO?
Yes. SAML 2.0 and OIDC are supported. LDAP and Active Directory integration is available on the Enterprise tier.

Does RowSpeak produce audit logs?
Yes. All user actions are logged locally — file uploads, AI queries, report generation, login/logout events, and admin changes. Logs are stored in your environment in JSON format, compatible with SIEM tools like Splunk and Elastic.

Compliance and Policy

Does RowSpeak help with GDPR compliance?
RowSpeak Private Deployment gives your organization full control over data handling, which supports your GDPR compliance posture. Since data stays in your environment, you remain the data controller. Your legal team should review the architecture against your specific requirements.

Does RowSpeak support SOX requirements?
RowSpeak's audit logging, access controls, and data isolation features support SOX-relevant controls. Your compliance team should review the technical architecture against your specific SOX requirements.

Can RowSpeak meet data residency requirements?
Yes. Since RowSpeak runs in your own infrastructure, data stays in whatever geographic location you choose. You control where the servers are.

Does RowSpeak have ISO 27001 or SOC 2 certification?
Contact us for the current status of our security certifications. We provide a detailed security architecture reference and data flow documentation to support your internal review process.

Can we run a penetration test on our RowSpeak deployment?
Yes. Since RowSpeak runs in your environment, you can run penetration tests and security assessments on your own schedule. Contact us if you need architecture documentation to support the assessment.

Models and AI

Which AI models does RowSpeak support?
Open-source: DeepSeek-V2, Qwen2.5, Kimi. Closed-source (via your API key): GPT-4o, Claude 3.5, Gemini 1.5 Pro. See the model selection guide for a full comparison.

Can we use different models for different workflows?
Yes. RowSpeak supports model routing — you can configure different models for different use cases within the same deployment.

Can we use a custom or fine-tuned model?
Custom model support is available on the Enterprise tier. Contact us to discuss your requirements.

Do we need to download model weights ourselves?
No. The Deployment Pack includes guidance on model acquisition. For air-gapped environments, we provide instructions for pre-loading model weights before deployment.

Performance and Reliability

What is the uptime SLA?
99.9% uptime target. See the performance benchmarks for full details.

How fast is the response time?
First token latency is under 100ms. Average full response time is 3–8 seconds depending on task complexity and hardware. See the performance benchmarks for configuration-specific data.

How many concurrent users can RowSpeak handle?
50+ concurrent users on a standard configuration (16-core, 64GB RAM, NVIDIA A10). Enterprise cluster configurations support 200+ concurrent users.

What happens if the server goes down?
RowSpeak recovers in under 60 seconds after a planned restart. For high-availability requirements, Kubernetes deployment with multiple nodes is recommended.

Updates and Support

How are updates delivered?
Updates are released as versioned Docker images or deployment packages. Your IT team applies updates on your schedule — there are no automatic updates.

Does RowSpeak provide remote support during deployment?
Yes. The RowSpeak team provides remote deployment support via video call and async communication. We help with environment setup, model configuration, and initial testing.

What support is available after go-live?
Support terms depend on your agreement. Contact us for details on post-deployment support options.

Can we roll back to a previous version?
Yes. Rolling back is done by reverting to the previous Docker image version. We provide rollback instructions with each release.

Evaluation and Procurement

What is the best first step? Book a demo. We will learn about your environment, use case, and requirements, and help you plan a structured evaluation.

Do you offer a proof-of-concept?
Yes. See the POC guide for a full walkthrough of what a RowSpeak POC involves.

Is there a free trial?
We offer a structured POC rather than a self-serve free trial for private deployment, since setup requires coordination with your IT environment. Contact us to discuss.

What documentation is available for procurement review?
The Deployment Pack includes reference architectures, hardware sizing guidance, model compatibility overview, and a deployment checklist. Request it here.

Still Have Questions?

Book a demo and we will answer your specific questions directly.

You can also review: